Rosmarinic acid is a novel inhibitor for Hepatitis B virus replication targeting viral epsilon RNA-polymerase interaction

Current therapeutics for hepatitis B virus (HBV) patients such as nucleoside analogs (NAs) are effective; however, new antiviral drugs against HBV are still desired. Since the interaction between the epsilon (ε) sequence of HBV pregenomic RNA and viral polymerase (Pol) is a key step in the HBV replication cycle, we aimed to identify small compounds for its inhibition, and established a pull-down assay system for the detection of ε-RNA-binding-Pol. Screening showed that 5 out of 3,965 compounds inhibited ε-Pol binding, and we identified rosmarinic acid, which exhibited specificity, as a potential antiviral agent. In order to examine the anti-HBV effects of rosmarinic acid, HBV-infected primary human hepatocytes from a humanized mouse liver were treated with rosmarinic acid. The rosmarinic acid treatment decreased HBV components including the amounts of extracellular HBV DNA with negligible cytotoxicity. We also investigated the combined effects of rosmarinic acid and the NA, lamivudine. rosmarinic acid slightly enhanced the anti-HBV activity of lamivudine, suggesting that the HBV replication step targeted by rosmarinic acid is distinct from that of NA. We analyzed an additional 25 rosmarinic acid derivatives, and found that 5 also inhibited ε-Pol. Structural comparisons between these derivatives implied that the “two phenolic hydroxyl groups at both ends” and the “caffeic acid-like structure” of rosmarinic acid are critical for the inhibition of ε-Pol binding. Collectively, our results demonstrate that rosmarinic acid inhibits HBV replication in HBV-infected cells by specifically targeting ε-Pol binding.


Introduction
Hepatitis B virus (HBV) infection is a major health issue worldwide, with approximately 248 million chronically infected individuals (CHB) [1]. Approximately 686,000 HBV-related deaths occur annually [2]. Interferon-α (IFN-α), pegylated IFN-α (PEG-IFN-α), and six nucleos(t)ide analogues (NAs), including lamivudine, entecavir, adefovir dipivoxil, tenofovir disoproxil fumarate, tenofovir alafenamide, and telbivudine, are currently approved for use in the clinical treatment of CHB patients [3]. Treatments with IFN have the potential to achieve HBsAg seroclearance by immunomodulation; however, not all patients respond to IFN. Although NAs more strongly suppress HBV replication than IFN by inhibiting reverse transcription (RT) with less side effects, the discontinuation of NAs may result in the relapse of HBV. Thus, life-long treatments with NAs are required, but may result in the emergence of resistant virus variants [4]. Since current therapeutics for CHB are insufficient, novel anti-HBV drugs are urgently required.
HBV Pol functions in many essential steps of HBV replication including RT, DNA synthesis, and RNA degradation. In addition to RT, the RNase H activity of Pol is a possible therapeutic target. Tavis and colleagues identified specific inhibitors for the RNase H activity of HBV Pol [18][19][20][21]. Besides its enzymatic roles, HBV Pol is crucially involved in pgRNA encapsidation. HBV Pol interacts with the ε sequence of pgRNA, and the ε-Pol interaction is an indispensable step for encapsidation [22]. Previous studies revealed that porphyrin compounds including hemin suppressed the ε-Pol interaction and subsequent protein-priming reaction [23]. Carbonyl J acid derivatives, known as HIV-1 RT inhibitors, have also been identified as ε-Pol binding and protein-priming inhibitors [24].
Since the interaction between ε and Pol is a distinct step from RT, ε-Pol binding inhibitors may be promising agents for combination therapy with NAs. However, large-scale screening to identify ε-Pol binding inhibitors has not yet been conducted. We herein established a screening system to search for ε-Pol binding inhibitors, and performed it using three chemical libraries including United States Food and Drug Administration (FDA)-, European Medicines Agency (EMA)-, and other agency-approved compounds. As a result of screening, rosmarinic acid and quercetin were identified as novel and specific inhibitors of ε-Pol binding. Quercetin has been reported to inhibit HBV replication. Therefore, we analyzed the anti-HBV effects of rosmarinic acid. Furthermore, an analysis of several types of rosmarinic acid derivatives revealed the critical structural features of rosmarinic acid for the inhibition of ε-Pol binding.

Compound treatment to cell lines stably expressing HBV
Hep38.7-Tet cells were seeded on collagen-coated 96-well plates at 1.0×10 4 cells/well. After cell attachment (2 to 3 h), culture medium was replaced with fresh FBS-free medium containing compounds 0, 1, 2, 3 days after seeding. FBS was added to culture medium 2 to 3 h after it had been changed. On day 5, extracellular HBV DNA was extracted from the culture supernatant and quantified by qPCR, as described below, and cells were then subjected to the WST-1 cell proliferation assay (Takara, MK400) following the manufacturer's instructions. CC50 in HepG2 cells was calculated from results at least 7 kinds of concentration of compound (Table 1).

Compound treatment to primary human hepatocytes infected with HBV
PXB-cells were infected at 5 Geq/cell using dHCGM containing serum from HBV-infected chimeric mice with humanized livers (PhoenixBio, PPC-BC), and 4% polyethylene glycol 8,000. Culture medium was replaced with fresh dHCGM containing compounds on days 1, 2, and 7 post-infection. On days 7 or 12, extracellular HBV DNA was extracted from the culture supernatant and quantified by qPCR, HBV RNA was extracted from cells and quantified by RT-qPCR, and SHBs were measured by ELISA as described below.

Electrophoresis mobility shift assay (EMSA)
EMSA was performed as described previously [29]. Briefly, a total of 30 pmol of the recombinant RIG-I protein was mixed with 5 pmol of synthetic dsRNA (25/25c) [29] and in a reaction mixture (20 mM Tris-HCl (pH 8.0), 1.5 mM MgCl 2 , and 1.5 mM DTT) in the presence of compounds. After an incubation at room temperature for 15 min, the reaction mixture was applied to a 15% acrylamide gel (TBE buffer) and dsRNA was detected by EtBr staining and RIG-I by Coomassie Brilliant Blue staining.

Helicase assay
The RIG-I helicase assay was performed as described previously [29].

Screening of ε-Pol inhibitors by a pull-down assay
In order to identify novel inhibitors of ε-Pol binding, we established an in vitro screening system based on a pull-down assay, as shown in Fig 1A. Briefly, chemically synthesized ε-biotin RNA was mixed with chemicals in the lysates of human cell lines expressing 3×FLAG-tagged-Pol, pulled down by streptavidin beads, and the ε-binding of 3×FLAG-Pol was then detected by Western blotting. As expected, 3×FLAG-Pol was pulled down with ε-biotin, but not with non-biotinylated ε or biotinylated RNA with an unrelated sequence (Fig 1B). Conditions were optimized further by using the previously reported inhibitors, hemin and calcomine orange 2RS [23,24] (Fig 1C), and we screened for novel inhibitors using chemical libraries. Primary screening was performed using compounds in a 10-drug mix, and candidate inhibitors were eventually identified by secondary screening using a single compound in the candidate 10-drug mix (S1A and S1B Fig). As a result, 5 out of 3,965 compounds were identified as inhibitors of ε-Pol binding. All 5 compounds: rosmarinic acid, quercetin, erythrosine B, merbromin, and verteporfin, exhibited inhibitory activities in dose-dependent manners (Fig 1D). Of note, verteporfin is a porphyrin, similar to hemin.

Rosmarinic acid and quercetin are specific inhibitors of ε-Pol
We examined the specificities of the 5 compounds for the inhibition of ε-Pol. In order to exclude the possibility that these compounds inhibit biotin-avidin binding, biotinylated DNA was pulled down by streptavidin sepharose in the presence or absence of the compounds. None of the 5 compounds affected binding between DNA-biotin and streptavidin sepharose (Fig 2A), suggesting that these compounds target the protein-RNA interaction. A previous study reported that RIG-I recognizes dsRNAs, such as the ε sequence in HBV pgRNA, and subsequently exhibits anti-HBV activity [30]. In order to examine whether the compounds identified affect RIG-I-dsRNA binding activity, we conducted an electrophoresis mobility shift assay (EMSA) to investigate the interaction between dsRNA and RIG-I in the presence or absence of the compounds. Rosmarinic acid and quercetin did not affect dsRNA-RIG-I binding, whereas the other compounds, erythrosin B, merbromin, and verteporfin, as well as the previously reported chemicals, hemin and calcomine orange 2RS, inhibited dsRNA-RIG-I binding (Fig 2B). We also tested whether the 5 compounds affect RIG-I helicase enzymatic activity, and found that only rosmarinic acid and quercetin did not affect the helicase activity of RIG-I, suggesting that RIG-I signaling is not influenced by treatments with rosmarinic acid and quercetin (S2A Fig). In order to further validate specificity, another RNP complex was tested. Liu and colleagues recently reported that ISG20 interacted with ε in an enzymatic activity (exonuclease)-independent manner [31]. 3×FLAG-tagged ISG20 D94G, an exonuclease inactive mutant, was pulled down with ε-biotin in the presence or absence of the compounds. Rosmarinic acid and quercetin did not affect ε-ISG20 binding, whereas erythrosin B, merbromin, verteporfin, hemin, and calcomine Orange 2RS exerted inhibitory effects (Fig 2C). These results suggest that rosmarinic acid and quercetin specifically inhibit the ε-Pol interaction.

Rosmarinic acid suppresses HBV replication
Quercetin was previously reported to inhibit HBV replication in HepG2.2.15 cells [32], a HepG2-derived cell line stably expressing HBV. Thus, we focused on the anti-HBV activity of rosmarinic acid using cell lines stably expressing HBV. No obvious cytotoxicity of rosmarinic acid was observed up to 30 μM in HepG2 cells and Hep38.7-Tet cells, a tetracycline inducible HBV-expressing cell line [27] (Fig 3B and 3D). The cytotoxicity of rosmarinic acid in HepG2 cells was also evaluated ( Table 1). Hep38.7-Tet cells were treated with rosmarinic acid, and the amount of HBV DNA secreted into the culture medium was quantified. The rosmarinic acid treatment significantly decreased extracellular HBV DNA with rough estimation of EC50 30 μM (Fig 3A). We then performed a combined treatment with rosmarinic acid and lamivudine. Although a low concentration of lamivudine (10 nM) exerted weak anti-HBV effects, the combined treatment enhanced inhibition (Fig 3C). PXB-cells, primary human hepatocytes isolated from an immunodeficient mouse with a humanized liver, were subjected to HBV infection and the effects of rosmarinic acid on its replication were examined. Rosmarinic acid showed no toxicity up to 100 μM (S3 Fig). Extracellular HBV DNA levels was significantly reduced and intracellular HBV RNA and extracellular HBsAg levels were slightly reduced by rosmarinic acid, suggesting that rosmarinic acid suppresses HBV replication in infected cells (Fig 4A-4C). In PXB-cells, the combined treatment with rosmarinic acid and lamivudine did not further enhance inhibition (Fig 4D-4F). Collectively, these results suggest that the rosmarinic acid treatment suppresses HBV replication in infected cells. Of note, HBV-infected PXB-cells were also treated with quercetin, and its inhibitory activity on HBV replication was confirmed (S4A-S4C Fig). We analyzed the rosmarinic acid derivatives listed in Table 2. Salvianolic acid A, Compound 2, 12c, 13c, and 21a [28] exhibited inhibitory activities against ε-Pol binding in dose-dependent manners (Fig 5A). Similar to rosmarinic acid, these 5 derivatives did not affect biotin-avidin binding, dsRNA-RIG-I binding, or RIG-I helicase activity (Fig 5B, 5C and S2B Fig). Furthermore, these derivatives inhibited HBV DNA production at similar levels to that of rosmarinic acid with low cytotoxicity (Fig 5D and Table 1). On the other hand, other derivatives did not exhibit inhibitory activities (Fig 5E).

Discussion
In the present study, we established a pull-down assay to screen for inhibitors of HBV ε-Pol binding. 3×FLAG-tagged full-length Pol was expressed in human cell lines, and cell lysates were used in the assay. It was not necessary to include an exogenous chaperone protein to detect the ε-Pol interaction. The ε-Pol interaction was detected in liver-derived (HepG2) and non-liver (HEK-293T) cells. Due to its high transfection efficiency, we used HEK-293T cells for screening. In order to test a larger number of chemicals, we mixed 10 compounds in primary screening. Although mixed chemicals may exhibit unexpected interference from each other, we identified 5 out of 3,965 chemicals with inhibitory activities. Rosmarinic acid and quercetin appeared to specifically inhibit the ε-Pol interaction. In Fig 4E, PXB cells treated with rosmarinic acid had significantly lower intracellular 3.5kb HBV RNA levels than DMSO-treated cells. Also, combined treatment with rosmarinic acid and lamivudine had lower 3.5kb HBV RNA level. This observation was unexpected if rosmarinic acid solely targets ε-Pol interaction. However, HBV progenies produced from PXB cells are known to have the potential to re-infect the cells. When rosmarinic acid specifically blocks the epsilon-polymerase binding, the treatment is estimated to reduce the production level of virus progenies and the rate of the secondary infection to PXB cells as well as to inhibit cccDNA synthesis from the pre-genome 3.5 kb HBV RNA. In such a context, we assume that HBV RNA levels in PXB cells were decreased by the treatment of rosmarinic acid.
Rosmarinic acid and quercetin are structurally related and both have two dihydoxybenzene moieties (Fig 6A). We tested the derivatives of rosmarinic acid and found that some did not exhibit inhibitory activities against the ε-Pol interaction. The examination with rosmarinic acid and its derivatives revealed a relationship between its chemical structure and ε-Pol inhibition (Figs 5 and 6). Compounds with inhibitory activities commonly had two dihydoxybenzene moieties ( Fig 6A); however, caffeic acid and its derivatives (C-4, -16a-k, -19, and -21b) did not exhibit inhibitory activity. Comparisons between C-12c and C-11a revealed that the two phenolic hydroxyl groups in the catechol structure were critical and the presence of 4 complete hydroxyl groups was required for inhibitory activity (compare C-12c with C-12a and C12d). Similarly, modifications to the hydroxyl groups into-OCH 3 groups (hydroxyl group masked with CH 3 ) resulted in the loss of inhibitory activity (compare C-12c with C-12b). On the other hand, the linker structure connecting the two catechol structures appeared to be Rosmarinic acid inhibits ε-Pol binding (D) Hep38.7-Tet cells were treated with 10 μM rosmarinic acid, compounds 12c, 13c, 2, and 21a, 100 nM lamivudine, or 400 ng/ml tetracycline. Extracellular HBV DNA and cell viability were measured as in (Fig 3A and 3B Rosmarinic acid inhibits ε-Pol binding flexible (compare C-12, C-2, C-13c, and C-21a). These results suggest that the 4 hydroxyl groups in the two catechol structures at both side chain ends are critical "active domains" for recognizing Pol. Whereas the linker structure was less critical for the specific interaction but the linker may determine critical distance between the two active domains. RIG-I is a viral RNA sensor for eliciting several antiviral responses [33,34]. Previous studies reported that innate immune signaling is important for anti-HBV responses [26,30]. ISG20 was shown to inhibit HBV replication in nuclease-dependent and -independent manners [31,35]. Since rosmarinic acid and quercetin inhibit ε-Pol binding without affecting dsRNA-RIG-I binding, RIG-I helicase activity, or ε-ISG20 binding (Figs 1D, 2B and 2C and S2A Fig), they presumably do not interfere with the host antiviral immune response.
Despite ε-Pol binding being strongly abolished by the rosmarinic acid treatment in vitro, the suppression of HBV replication in cells by rosmarinic acid was less efficient than the nucleotide analogue, lamivudine. One of the reasons for insufficient inhibition is permeability. Another potential reason is chemical stability in cells. These limitations may be overcome in future studies. Rosmarinic acid is a natural compound (abundant in Lamiaceae herbs including spearmint, sage, peppermint, and perilla) and is utilized as a dietary supplement as well as Chinese herbal medicine. To date, no detrimental effects on humans have been reported, suggesting lower toxicity in vivo. Therefore, if we optimize drug administration, we may be able to evaluate the anti-HBV effects of rosmarinic acid in vivo.